measurement <- c(0.704,  0.685,  0.895,  0.377,  0.711,  0.355)
 
       #Each position contains the methylation value for BS74, BS71, BS174, BS134, ML15, and BS22, respectively
                                                                 #When a methylation value is not available, its position should be substituted by NaN

combinations <- combn(seq(6), 2)

# the regression coefficient for each pair of measures
x <- list()
x[[1]]  <- c(-0.23699083,  0.31230647, 0.429878837188)
x[[2]]  <- c(-0.28109206,  0.21857603, 0.474067456431)
x[[3]]  <- c(-0.25389859, -0.2362705,  0.750660923371)
x[[4]]  <- c(-0.17378906, -0.35035923, 0.793634538072)
x[[5]]  <- c(-0.21169592, -0.16500069, 0.634663208437)
x[[6]]  <- c( 0.26912685,  0.20000115, 0.157248443428)
x[[7]]  <- c( 0.27919318, -0.25606219, 0.448793002983)
x[[8]]  <- c( 0.24637673, -0.3495989,  0.543116236858)
x[[9]] <- c( 0.19319181, -0.21946734, 0.427286215609)
x[[10]] <- c( 0.23783653, -0.2739245,  0.437915653474)
x[[11]] <- c( 0.1568182,  -0.35084666, 0.559689788194)
x[[12]] <- c( 0.23990307, -0.19578163, 0.344194030962)
x[[13]] <- c(-0.11983331, -0.34994783, 0.743168183041)
x[[14]] <- c(-0.252382,   -0.20448561, 0.67323140595)
x[[15]] <- c(-0.31900696, -0.13367664, 0.694656738874)

coefficients <- x

predictions <- c()

# for each pair of measured methylation levels
for (i in 1:ncol(combinations)){
    # get the current pair
    curr_combination = combinations[,i]
    m1 <- measurement[curr_combination[1]]
    m2 <- measurement[curr_combination[2]]

    # if one of the level is nan (not measured), ignore it
    if (is.nan(m1) || is.nan(m2)) next

    # get the regression coefficient for this pair
    curr_coefficient = coefficients[[i]]
    c1 <- curr_coefficient[1]
    c2 <- curr_coefficient[2]
    intercept <- curr_coefficient[3]
    
    # predict an age
    curr_prediction <- c1 * m1 + c2 * m2 + intercept
    predictions <- c(predictions, curr_prediction)
}

# get the average of all predictionss
mean_prediction <- mean(predictions)

# and adjust the average by an additional regression
final_prediction <- 2.34681774 * mean_prediction - 0.560280844213

print(final_prediction)